High Efficiency Video Coding (HEVC) is an evolving video compression standard. It is under development by the Joint Collaborative Team on Video Coding (JCT-VC) formed by ISO/IEC Moving Picture Experts Group (MPEG) and ITU-T Video Coding Experts Group (VCEG). HEVC can employ several modules such as intra/inter prediction, transform, quantization and in-loop filtering and entropy coding modules.
As the standardization of HEVC is reaching completion, JCT-VC has started planning to add a Scalable Video Coding (SVC) extension into the HEVC standard. SVC is a way of coping with the heterogeneity of networks and devices in modern video service environments. A SVC bit stream may contain several subset bit streams that can themselves be decoded, where these sub streams may each represent the source video content but with different respective resolutions, frame rates, quality, and/or bit depths, for example. On the decoding side, a particular subset stream may be chosen for decoding, where the choice may be based on a desired resolution, frame rate, quality, and/or bit depth, etc.
The scalability is achieved by using a multi-layer coding structure. In general, there may be one base layer (BL) and several enhancement layers (ELs) in a SVC system. The BL may be encoded in a manner that is compatible with the HEVC standard, and may be identified as layer 0. ELs may be identified as layers 1, 2, . . . etc. When coding an EL with layer identifier equal to N, all the layers with a layer identifier less than N are available. This means that the picture corresponding to an EL may be predicted from lower layer pictures (i.e., layers having a lesser number) or from previously coded pictures in the same layer. All the layers may be sent to the decoding side, where a particular layer may be chosen for decoding.
Because each layer represents the source video content, the transmission and storage of multiple layers represents a significant amount of data. While an end user is provided with several options as to which representation (i.e., layer) to decode, significant bandwidth and processing power may be consumed in the encoding, transmission, and storage of these layers.
In the drawings, the leftmost digit(s) of a reference number identifies the drawing in which the reference number first appears.